Prediction of Socioeconomic Levels Using Cell Phone Records

Abstract

Abstract. The socioeconomic status of a population or an individual provides an understanding of its access to housing, education, health or basic services like water and electricity. In itself, it is also an indirect indicator of the purchasing power and as such a key element when personalizing the interaction with a customer, especially for marketing campaigns or offers of new products. In this paper we study if the information derived from the aggregated use of cell phone records can be used to identify the socioeconomic levels of a population. We present predictive models constructed with SVMs and Random Forests that use the aggregated behavioral variables of the communication antennas to predict socioeconomic levels. Our results show correct prediction rates of over 80% for an urban population of around 500,000 citizens.

Forecasting socioeconomic trends with cell phone records

“…Previous research using cell phone calling logs has already shown that cell phone-based behavioral patterns are correlated to specific socio-economic characteristics [3,11]. For example, Eagle et al showed correlations between the size of a cell phone social network and the socioeconomic level of a person, and Frias et al observed strong relationships between mobility and socio-economic indices [6].…”

“…For example, Eagle et al showed correlations between the size of a cell phone social network and the socioeconomic level of a person, and Frias et al observed strong relationships between mobility and socio-economic indices [6]. Additionally, Soto et al showed that the socioeconomic level(SEL) of a region at a given moment in time, can be predicted from cell phone activity during the same time period with an 80% of accuracy using training sets containing both SELs and calling activity [11,4]. Although this work threw some light onto the predictability of SELs using calling data, it focused on predicting the present i.e., SEL values at a specific moment in time rather than forecasting future values.…”

Computational Sustainability and Artificial Intelligence in the Developing World

“…We have empirically evaluated CenCell with millions of cell phone records from urban citizens and we have shown that it correctly determines the socioeconomic levels computed by the NSIs with high accuracies (Soto et al 2011). Thus, CenCell significantly decreases the workload of the enumerators that carry out the interviews and as such, allows us to reduce the budget allocated for the computation of census maps.…”

ICT4ICTD: Computational Social Science for Digital Development

“…Using information from mobile phone call records through a plurality of base stations, it has been shown how the prediction of socioeconomic level in a geographic region can automatically be performed [24,25]. Prediction accuracy depends on the kind of variable (for example, predicting gender from mobile phone behavior is surprisingly tricky [26,27]), but generally it has been shown to be around 80-85% when using mobile call data records like call duration or frequency [24,[26][27][28]. Such ideas can be fine-tuned for cases where more detailed digital footprints are available.…”